SAIL

KO | EN

gitlite — search

#sql #go #artificial-intelligence #claude #gemini #pdf #codex #deep-learning #python #json

(★ 9)

No description available.

File Explorer

ZIP Download

- - theme_overrides.css
- - robomimic.algo.rst
  - robomimic.config.rst
  - robomimic.envs.rst
  - robomimic.models.rst
  - robomimic.rst
  - robomimic.utils.rst
- - d4rl.md
  - mimicgen.md
  - momart.md
  - overview.md
  - robomimic_v0.1.md
  - robosuite.md
  - roboturk_pilot.md
- - can.png
  - can_paired.png
  - can_real.jpg
  - core_features.png
  - lift.png
  - lift_real.jpg
  - machine_generated.png
  - mimicgen_mosaic.gif
  - module_overview.png
  - modules.png
  - momart_bowl_in_sink.png
  - momart_dump_trash.png
  - momart_grab_bowl.png
  - momart_open_dishwasher.png
  - momart_open_dresser.png
  - momart_table_cleanup_to_dishwasher_overview.png
  - momart_table_cleanup_to_sink_overview.png
  - momart_table_setup_from_dishwasher_overview.png
  - momart_table_setup_from_dresser_overview.png
  - momart_unload_dishwasher_to_dresser_overview.png
  - multi_human.png
  - proficient_human.png
  - randomizer.gif
  - roboturk_pilot.png
  - square.png
  - task_can.gif
  - task_can_real.gif
  - task_lift.gif
  - task_lift_real.gif
  - task_square.gif
  - task_tool_hang.gif
  - task_tool_hang_real.gif
  - task_transport.gif
  - tensorboard.png
  - tool_hang.png
  - tool_hang_real.jpg
  - transport.png
- - getting_started.md
  - implemented_algorithms.md
  - installation.md
  - overview.md
- - acknowledgments.md
  - contributing.md
  - references.md
  - team.md
  - troubleshooting.md
- - robomimic_v0.1.md
- - algorithms.md
  - configs.md
  - dataset.md
  - environments.md
  - models.md
  - overview.md
- - configs.md
  - custom_algorithms.md
  - dataset_contents.md
  - hyperparam_scan.md
  - observations.md
  - pretrained_representations.md
  - reproducing_experiments.md
  - tensor_collections.md
  - training_transformers.md
  - using_pretrained_models.md
  - viewing_results.md
- - v0.2.md
- .gitignore
- conf.py
- index.rst
- Makefile
- robomimic_logo.png
- - datasets.ipynb
  - run_policy.ipynb
- add_new_modality.py
- simple_config.py
- simple_obs_nets.py
- simple_train_loop.py
- train_bc_rnn.py
- - __init__.py
  - act.py
  - algo.py
  - bc.py
  - bcq.py
  - cql.py
  - diffusion_policy.py
  - gl.py
  - hbc.py
  - iql.py
  - iris.py
  - td3_bc.py
- - __init__.py
  - act_config.py
  - base_config.py
  - bc_config.py
  - bcq_config.py
  - config.py
  - cql_config.py
  - diffusion_policy_config.py
  - gl_config.py
  - guide_config.py
  - hbc_config.py
  - iql_config.py
  - iris_config.py
  - td3_bc_config.py
- - __init__.py
  - env_base.py
  - env_gym.py
  - env_ig_momart.py
  - env_real_panda.py
  - env_real_panda_gprs.py
  - env_robosuite.py
  - wrappers.py
- - __init__.py
  - base_nets.py
  - distributions.py
  - guiding_modules.py
  - obs_core.py
  - obs_nets.py
  - policy_nets.py
  - transformers.py
  - vae_nets.py
  - value_nets.py
- - __init__.py
  - default_baxter.json
  - default_iiwa.json
  - default_jaco.json
  - default_kinova3.json
  - default_panda.json
  - default_sawyer.json
  - default_ur5e.json
  - ik_pose.json
  - joint_position.json
  - joint_position_SAIL.json
  - joint_torque.json
  - joint_velocity.json
  - joint_velocity_SAIL.json
  - osc_pose.json
  - osc_pose_SAIL.json
  - osc_position.json
  - paths.py
- - __init__.py
  - eval_suite.py
  - installation.sh
  - run_trained_agent_receding_horizon.py
  - USAGE.md
- - add_relative_eef.py
  - convert_actions.py
  - dataset_states_to_obs.py
  - download_datasets.py
  - download_momart_datasets.py
  - extract_obs_from_raw_datasets.sh
  - extract_obs_from_raw_datasets_v2.sh
  - generate_config_templates.py
  - generate_paper_configs.py
  - get_dataset_info.py
  - give_slack_notification.py
  - hyperparam_helper.py
  - hyperparam_helper_diffusion.py
  - playback_dataset.py
  - postprocess_dataset_intervention_segments.py
  - remove_idle_segments.py
  - run_trained_agent.py
  - setup_macros.py
  - split_train_val.py
  - train.py
- - __init__.py
  - action_utils.py
  - dataset.py
  - deco_utils.py
  - env_utils.py
  - file_utils.py
  - hyperparam_utils.py
  - log_utils.py
  - loss_utils.py
  - obs_utils.py
  - python_utils.py
  - ros_utils.py
  - script_utils.py
  - tensor_utils.py
  - test_utils.py
  - torch_utils.py
  - train_utils.py
  - transform_utils.py
  - vis_utils.py
- __init__.py
- macros.py
- test.sh
- test_bc.py
- test_bcq.py
- test_cql.py
- test_examples.py
- test_hbc.py
- test_iql.py
- test_iris.py
- test_scripts.py
- - README.md
  - robosuite-v1.4.1-sail.patch
- apply_robosuite_patch.sh
.gitignore
CITATION.cff
environment.yml
LICENSE
MANIFEST.in
NOTICE
README.md
requirements-docs.txt
requirements.txt
setup.py

# Use via CDN

jsDelivr

jsDelivr serves any public GitHub repository as a CDN with zero setup. Pick a version and a file to get a ready-to-paste link and snippet.

Version

File

Link

Example

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

[**Project page**](https://nadunranawaka1.github.io/sail-policy) · [**Paper (arXiv:2506.11948)**](https://arxiv.org/abs/2506.11948)

This repository is the companion code for **SAIL** (*Speed Adaptation for Imitation Learning*): a full-stack system for executing visuomotor imitation-learning policies **faster than demonstration speed** while preserving task success.

It is built on [robomimic](https://github.com/ARISE-Initiative/robomimic) and adds dataset processing, training configs, Error-Adaptive Guidance (EAG) evaluation, and adaptive speed modulation used in the paper.

## Release timeline

| Phase | What | Status |
|:-----:|------|--------|
| **1** | **Simulation** — training, dataset processing, EAG evaluation, and the robosuite SAIL patch | ![Released](https://img.shields.io/badge/Released-2ea44f?style=flat-square) |
| **2** | **Real hardware** — Franka (Deoxys) and UR5 (ROS&nbsp;2) stacks from the paper | ![Coming soon](https://img.shields.io/badge/Coming_soon-6e7781?style=flat-square) |

## Installation

Requires [conda](https://docs.conda.io/). From the repository root:

```bash
bash robomimic/SAIL/installation.sh
conda activate SAIL
```

The install script will:

1. Create/reuse a conda env (`SAIL` by default) with PyTorch 2.1
2. Clone [ARISE robosuite](https://github.com/ARISE-Initiative/robosuite) at a pinned `v1.4.1` commit and apply [`third_party/patches/robosuite-v1.4.1-sail.patch`](third_party/patches/robosuite-v1.4.1-sail.patch)
3. Install this package editable
4. Install stock [AWE](https://github.com/lucys0/awe) for waypoint extraction

**Important:** apply the robosuite patch only against the pinned commit (`b9d8d3de`). See [`third_party/patches/README.md`](third_party/patches/README.md).

Environment overrides:

| Variable | Meaning |
|----------|---------|
| `SAIL_ENV_NAME` | Conda env name (default `SAIL`) |
| `SAIL_PYTORCH_CUDA` | CUDA build for conda pytorch (default `11.8`) |
| `ROBOSUITE_INSTALL_DIR` | Where to clone/patch robosuite |
| `AWE_INSTALL_DIR` | Where to clone AWE |

## Dataset processing

Create reached-pose actions and adaptive-speed labels:

```bash
# Reached actions
python robomimic/SAIL/dataset_processing/add_all_actions.py --dataset=<PATH_TO_DATASET>

# AWE waypoints + precision / speed labels
python robomimic/SAIL/precision_processing/save_awe_waypoint_concurrent.py --dataset=<PATH_TO_DATASET>
python robomimic/SAIL/precision_processing/label_awe_trajectory_precision.py --dataset=<PATH_TO_DATASET>
```

Tips:

- Tune `--err_threshold` in `save_awe_waypoint_concurrent.py` to match waypoint density from the AWE paper.
- Tune clustering parameters in `label_awe_trajectory_precision.py` for precise vs non-precise balance.
- Defaults work well for robomimic Lift / Can / Square.
- For repetitive pick-and-place demos, label sub-tasks separately if needed.

Visualize speed labels (control-frequency graph + agentview overlay):

```bash
python robomimic/SAIL/precision_processing/dynamic_speed_awe.py \
  --dataset=<PATH_TO_DATASET> --video_path=<OUTPUT.mp4>
```

## Training

Example Diffusion Policy config with EAG conditioning:

[`robomimic/exps/templates/diffusion_policy_SAIL.json`](robomimic/exps/templates/diffusion_policy_SAIL.json)

```bash
python robomimic/scripts/train.py --config=robomimic/exps/templates/diffusion_policy_SAIL.json
```

Edit the config `train.data` path to point at your processed dataset.

## Evaluation (receding horizon + EAG)

```bash
python robomimic/SAIL/run_trained_agent_receding_horizon.py \
  --agent=<PATH_TO_CHECKPOINT.pth> \
  --guide_config=robomimic/SAIL/guide_template/base_cfg_weight_1.json
```

Example guide configs are under [`robomimic/SAIL/guide_template/`](robomimic/SAIL/guide_template/). Controller JSONs for high-fidelity tracking are under [`robomimic/robosuite_configs/*_SAIL.json`](robomimic/robosuite_configs/).

## Citation

If you use this code, please cite SAIL:

```bibtex
@article{arachchige2025sail,
  title={SAIL: Faster-than-Demonstration Execution of Imitation Learning Policies},
  author={Arachchige, Nadun Ranawaka and Chen, Zhenyang and Jung, Wonsuhk and Shin, Woo Chul and Bansal, Rohan and Barroso, Pierre and He, Yu Hang and Lin, Yingyan Celine and Joffe, Benjamin and Kousik, Shreyas and Xu, Danfei},
  journal={arXiv preprint arXiv:2506.11948},
  year={2025}
}
```

This repository extends robomimic; please also cite:

```bibtex
@inproceedings{robomimic2021,
  title={What Matters in Learning from Offline Human Demonstrations for Robot Manipulation},
  author={Ajay Mandlekar and Danfei Xu and Josiah Wong and Soroush Nasiriany and Chen Wang and Rohun Kulkarni and Li Fei-Fei and Silvio Savarese and Yuke Zhu and Roberto Mart\'{i}n-Mart\'{i}n},
  booktitle={Conference on Robot Learning (CoRL)},
  year={2021}
}
```

See [`CITATION.cff`](CITATION.cff) and [`NOTICE`](NOTICE) for third-party attributions (robosuite, AWE, Diffusion Policy, etc.).

## Acknowledgments

This work was supported by the State of Georgia and the Agricultural Technology Research Program at Georgia Tech; the AI Manufacturing Pilot Facility / Georgia AIM (U.S. Department of Commerce EDA Award 04-79-07808); NSF CCF Award 2211815; and NSF Award 1937592.

We thank the ARISE Initiative for [robomimic](https://github.com/ARISE-Initiative/robomimic) and [robosuite](https://github.com/ARISE-Initiative/robosuite), and the AWE authors for [Automatic Waypoint Extraction](https://github.com/lucys0/awe).

## License

MIT — see [`LICENSE`](LICENSE). Upstream robomimic remains copyright Stanford Vision and Learning Lab; SAIL modifications are copyright Georgia Institute of Technology / the SAIL authors.

SAIL

File Explorer

# Use via CDN

Was this content helpful?

Recent Feedback

Download README